Variation in Detection of Adenomas and Polyps by Colonoscopy and Change Over Time With a Performance Improvement Program

CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2009;7:1335 1340 Variation in Detection of Adenomas and Polyps by Colonoscopy and Change Over Time With a Performance Improvement Program AASMA SHAUKAT,*, CRISTINA OANCEA, JOHN H. BOND, TIMOTHY R. CHURCH, and JOHN I. ALLEN, *Department of Medicine, Minneapolis Veterans Affairs Medical Center; Division of Gastroenterology, Department of Medicine, and Department of Environmental Health Sciences, School of Public Health, University of Minnesota; and Minnesota Gastroenterology PA, Minneapolis, Minnesota BACKGROUND & AIMS: There has been no prospective, community-based study to track changes in adenoma detection by individual physicians over time and to determine the effectiveness of targeted educational interventions. METHODS: We prospectively collected information on 47,253 screening colonoscopies in average-risk individuals 50 years and older performed by a community-based practice in the Twin Cities of Minnesota. During a period of 3 years, 5 specific interventions were implemented; each was designed to improve adenoma detection rates. Controlling for patient-related and procedurerelated factors, rates of adenoma detection and 3-year trends for individual physicians were plotted, and intraclass correlation coefficients were calculated. Generalized estimating equations were used to identify factors associated with detection of adenomas and polyps. RESULTS: At least 1 polyp and 1 adenoma were found in 36% and 22% of examinations, respectively. Adenoma detection rates by endoscopists ranged from 10% 39%. There was no significant improvement during the study period despite planned, systematic interventions. Factors associated with adenoma detection included age of the patient (odds ratio [OR], 1.02; 95% confidence interval [CI], 1.02 1.02), male sex (OR, 1.53; 95% CI, 1.34 1.74), and adequate preparation quality (OR, 2.26; 95% CI, 1.64 3.12). CON- CLUSIONS: The detection of adenomas by individual physicians during a 3-year period varied and did not appear to change between individual endoscopists, despite planned, systematic interventions. This indicates that other targeted interventions might be required to improve adenoma detection rates among experienced, community gastroenterologists. Most colorectal cancers (CRCs) arise from neoplastic polyps including adenomas and recently described sessile serrated adenomas. 1,2 During the past decade, colonoscopy has become the screening test of choice, because it allows for both detection and removal of asymptomatic adenomas. 3,4 Successful cancer reduction occurs when all visible polyps are detected and removed during a high quality initial examination and subsequent, appropriately timed surveillance examinations are performed. Variability in the quality of colonoscopy might explain in part the published miss rates for CRC of 0.5% 5% 5,6 and 6% 18% for adenomas 5 mm 7. Multiple studies have assessed variability in colonoscopic practice and demonstrated variation in adenoma find rate for average-risk individuals undergoing screening colonoscopy. Barclay et al 8 documented variability in detection rates of both small and advanced adenomas among 12 endoscopists. An average withdrawal time of less than 6 minutes was associated with lower detection rates (28.3% vs 11.8%), underscoring the importance of careful endoscopic examination. A more recent article by Barclay et al 9 demonstrated a significant increase in adenoma detection among experienced colonoscopists by using an audible timer to ensure at least 2 minutes of inspection time for each of 4 colonic segments during screening examinations. Others have emphasized the relationship between adenoma detection at colonoscopy and factors such as preparation quality, level of sedation, and total procedure time. 10,11 Withdrawal time measurement has become a common surrogate measure for colonoscopy quality in numerous quality improvement programs nationwide, but a recent study has suggested that factors other than increasing withdrawal time will be needed to improve adenoma detection among those endoscopists who have low baseline detection rates. 12 Each of the above studies suggests that improving an individual endoscopist s adenoma detection can be achieved by modifying operator technique or improving preparation quality. What is not clear is how much techniques such as feedback and education can alter detection rates. For colonoscopy to be an effective screening tool, in the face of alternative means of total colonic examination, improvement of polyp detection for low performing endoscopists must be feasible and sustainable over time. 13 15 Before any effective quality assurance program can be developed and implemented, the variability in adenoma detection among physicians and effect of educational and quality improvement programs on detection rates over time need to be studied. Also, factors associated with adenoma and polyp detection rates need to be further studied. The aim of our study was to systematically evaluate variability in and compare rates of adenoma detection over time for individual physicians after application of specific quality improvement programs. We also sought to define factors associated with detection of adenomas and polyps in an average-risk screening population of men and women older than 50 years. Methods Record Abstraction We collected information on all colonoscopy examinations performed within 5 ambulatory endoscopy centers (AECs) owned by a 51-physician community-based practice in the Twin Abbreviations used in this paper: AEC, ambulatory endoscopy center; CI, confidence interval; CRC, colorectal cancer; GEE, generalized estimating equations; OR, odds ratio. 2009 by the AGA Institute 1542-3565/09/$36.00 doi:10.1016/j.cgh.2009.07.027

1336 SHAUKAT ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 7, No. 12 Cities of Minnesota (Minnesota Gastroenterology) during a 3-year period from January 2004 to December 2006. Information including patient demographics, endoscopic findings, and pathology results are contained within a single database of the electronic medical record (NextGen Healthcare, Atlanta, GA). Only first-time screening colonoscopies performed in averagerisk individuals 50 years and older were included in this report. Of all patients seen in the AECs, 87% were directly scheduled for colonoscopy through an open access system, with predefined clinical algorithms for triage and pre-procedure evaluation. Although formal analysis has not been performed on patient demographics among partners, there are likely few differences among first-time screening patients in this type of open referral system except for an increased number of female patients seen by female partners. All colonoscopies were performed during 30-minute time slots by Board Certified gastroenterologists, each of whom had performed more than 5000 examinations before the beginning of the study period. Patients underwent standard bowel preparation by using one of several oral lavage regimens. All procedures were performed by using moderate conscious sedation with midazolam and fentanyl. Endoscopists were not aware of the specific study hypothesis, but all partners had previously signed a partnership agreement that included an acknowledgment that results of their procedures will be monitored for quality purposes. We identified all notes with the term screening in the indications field (from a drop-down menu), in addition to a detailed algorithm that uses text word search for all words resembling screening or its synonyms that could have been entered as free text in the indication field. As part of the routine practice, each patient was asked specifically if the current examination was their first colonoscopy, and only those data were entered into this report. From the procedure note we extracted information on quality of preparation (defined as adequate including descriptions of good, excellent, or fair, or inadequate, including descriptions of poor), completeness of procedure, location and number of polyps, conscious sedation administered (type and doses), and total duration of procedure (defined as time from insertion of colonscope to removal from anal verge). For this study, all polyps that were tubular adenoma, villous adenoma, tubulovillous adenoma, or sessile serrated adenoma were labeled as adenomas. Colonoscopy was defined as complete if cecal landmarks or surgical anastomosis was reached and documented. Visual confirmation of the endoscopists declaration of cecal intubation is expected as part of the responsibilities of the endoscopy nurses employed by the practice. Educational Interventions and Feedback We correlated the effects of specific quality improvement interventions with changes in adenoma detection among all endoscopists during the 3-year time period. Five specific interventions were implemented; each was designed to improve adenoma detection rates. In order, they included (1) review of individual adenoma rates (blinded), (2) review of rates unblinded within partnership meetings, (3) education about published colonoscopy quality literature including definition of expected sex-specific adenoma rates and the importance of a slow withdrawal time, (4) personal discussions between practice leaders and poorly performing endoscopists, and (5) financial consequences for not achieving a 6-minute withdrawal time in 95% of colonoscopy examinations; specifically, 1% of each partner s total compensation was at risk if they did not achieve 6-minute withdrawal time in 95% of screening colonoscopy examinations each year. During the calendar year 2004, data were collected without anyone s knowledge that adenoma rates were being monitored (excluding one of the authors, J.I.A.). In January 2005, adenoma detection rates for individual physicians were presented at a group retreat. The data were blinded so individual physicians did not know their own or their partners detection rate. This presentation included a discussion of pertinent literature including the importance of the adenoma find rate. Beginning in June 2005, each physician began receiving their own data (completion rate and sex-specific adenoma detection rates), with benchmarking to the group as whole and published data. Every 6 months thereafter, updated data were mailed to individual physicians. In January 2006, physician names were attached to adenoma find rates, and these unblinded data were discussed among the partners at a group meeting. In June 2006, letters were sent to all physicians in the practice with unblinded adenoma find rates. Personal discussions between practice leaders and individual partners (those with low adenoma rates) began in 2006. Discussions regarding financial incentives began in June 2006. Statistical Analysis We categorized location of polyp as right-sided if it was located in the cecum, ascending colon, hepatic flexure, or transverse colon or left-sided if it was located in the rectum, sigmoid colon, descending colon, or splenic flexure. Our first aim was to calculate adenoma detection rates for each physician, for an average patient who presents for colonoscopy to the practice; plot the adenoma rates over time for each physician; and compare the variability of all physicians over time. For this purpose, we restricted the analysis to physicians who had performed at least 100 screening colonoscopies during each of the 3 years of study and also excluded individuals with missing or incomplete information on performing physician. Forty-three gastroenterologists met this criterion and were included. We used generalized linear mixed models with random Table 1. Baseline Characteristics of the Participants and Distribution of Procedure-Related Factors a Baseline variables n(%) or mean SD, N 47,253 Age (y) 59.3 8 Sex Male 21,818 (46%) Female 25,396 (54%) Preparation quality adequate 46,738 (99.1%) Completed procedures (%) 46,292 (98%) Total duration of procedure (min) 19 21 Dose of fentanyl ( g) 136 140 Dose of midazolam (mg) 2.6 1 Polyp detection 16,796 (36%) Adenoma detection 10,402 (22%) At least 1 left-sided polyp 11,175 (23.6%) At least 1 right-sided polyp 8040 (17%) SD, standard deviation. a Missing information for sex 39 (0.08%), preparation quality 133 (0.2%).

December 2009 VARIABILITY IN ADENOMA DETECTION AND CHANGE OVER TIME 1337 intercept and time nested within physicians entered in the model as a covariate to study the variability in outcome by physician over time. Data were adjusted for mean age for patients, frequency of male patients, and frequency of adequate preparation quality from the total sample. The intraclass correlation coefficient was used to determine the correlation among patients within clinician clusters. Our secondary aim was to study factors associated with adenoma and polyp detection. For this purpose, categorical and continuous variables were compared by using the 2 and Student t test, respectively. We used the Fisher exact test and log-rank test for categorical and continuous data, respectively, when the data were not normally distributed. The dependent variable was adenoma detection. The independent variables of interest were patient age, sex, preparation quality, sedation administered, and average duration of procedure for examinations where no polyps were found. We used generalized estimating equations (GEE) to identify factors associated with detection of adenoma. This allows analysis of variables of interest clustered by physician. We performed stratified analysis in men and women separately and also looked for any interaction between sex and age, adequate preparation quality, or duration of procedure. We performed similar analysis for detection of polyps. Results During a period of 3 years, a total of 97,623 colonoscopy examinations were performed in the 5 AECs by 51 gastroenterologists, of which 47,253 were screening examinations. Ninety-eight percent of the examinations were complete. Mean age of patients was 59.3 8 years. Characteristics of the study Figure 1. Patterns of detection of adenoma during the 3 years plotted for individual physicians adjusted for age, sex, and adequate preparation quality, which are split in 4 panes by quartiles of procedure volume for ease of viewing. Each line represents one physician. The quartile cutoffs were as follows: (A) 100 870 colonoscopies during study duration, (B) 871 1140 colonoscopies during study duration, (C) 1141 1335 colonoscopies over study duration, and (D) more than 1335 colonoscopies during study duration. Green arrows: adenoma detection rates for individual physicians were presented at a group retreat. Blue arrows: physicians received their own adenoma detection rate, with benchmarking to the group as a whole and published data. Red arrows: physician names were attached to adenoma detection rates, and these unblinded data were discussed at group meeting. Yellow arrows: letters sent to all physicians in practice with unblinded adenoma detection rates, and personal discussion with practice leaders and individual physicians were held.

1338 SHAUKAT ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 7, No. 12 Table 2. Distribution of Patient and Procedure-Related Factors by Adenoma Detection a Adenoma detected, n 10,402 Adenoma not detected, n 36,851 Age (y) 60 8 59 8 Sex Male 5973 (57%) 15,845 (43%) Female 4422 (43%) 20,974 (57%) Adequate preparation quality 10,333 (99.5%) 36,405 (99.1%) Completion rate 10,313 (99.1%) 35,979 (97.6%) Duration of procedure a (min) 17.3 13 19.7 22 Dose of fentanyl ( g) 142 172 135 138 Dose of midazolam (mg) 2.5 1 2.6 1 Adenoma location Left-sided 3883 (37.3%) Right-sided 3807 (36.6%) Both 2380 (23%) Unknown 332 (3%) a Missing information for sex 39 (0.08%), preparation quality 133 (0.2%). population and procedure-related factors are presented in Table 1. At least 1 polyp was found in 36% of examinations and at least 1 adenoma in 22% of examinations (17.4% among women, 27.3% among men). Preparation quality was considered adequate in the majority (99.1%) of the examinations. Adenoma detection rates per physicians adjusted for age, sex, and adequate preparation quality varied from 10% 39%. Patterns of detection of adenoma during the 3 years plotted for individual physicians are shown in Figure 1 (split in 4 panes by quartiles of procedure volume for ease of viewing). As can be appreciated visually, there were no secular patterns or trends over time, despite the educational programs and feedback. The variation for an individual physician relative to variation for the group was random, suggesting that physicians tend to have a relatively constant rate of adenoma detection, and the applied interventions were ineffective in changing rates during the 3-year study period. The intraclass correlation coefficient for variability in adenoma detection over time was 0.009. Odds of detecting adenoma during a procedure were higher with increasing age (odds ratio [OR], 1.022; 95% confidence interval [CI], 1.020 1.025), being male (OR, 1.532; 95% CI, 1.347 1.741), adequate quality of adequate preparation (OR, 2.269; 95% CI, 1.647 3.126), and shorter total duration of procedure (where no polyp was found) (OR, 0.979; 95% CI, 0.961 0.996). When stratified by sex, the results were similar (data not shown), but we did find an interaction between sex and duration of procedure, where increasing duration of procedure was associated with a greater decrease in adenoma find rate for women than for men (OR, 1.009; 95% CI, 1.002 1.015) (Tables 2 and 3). Factors associated with polyp detection were similar to those found for adenoma detection and included age of the patient, male sex, adequate preparation quality, and shorter duration of procedure (data not shown). We found similar interaction between sex and duration of procedure as for adenoma detection. Dose of midazolam or fentanyl and completion rate were not associated with detection of adenoma or polyps. We also did not find differences in factors associated with right-sided or left-sided adenomas or polyps. Discussion In this large community-based study of 43 Board Certified gastroenterologists who performed 47,253 screening colonoscopies, adenoma detection rates for individual physicians varied from 10% 39% and did not change in response to 5 separate educational and feedback interventions. Failure of education efforts to change adenoma detection might have several possible explanations. First, educational programs or feedback used might not have been compelling enough to alter behavior. Of note, despite emphasis on improving adenoma detection rates as a goal of colonoscopy quality programs, there are no studies that have defined the type of interventions, applied when, and how long they would be effective. Second, it might be difficult to improve adenoma detection rates for colonoscopists once they are proficient in their skill and have developed a comfort level or set of beliefs on internal tradeoffs between spending more time examining the colon or changing techniques and its yield. The limitations of these results were that there might be residual confounding in calculation of adenoma detection rate over time that might be contributing to the variability and lack of secular trends over time that we were unable to account or control for, such as patient diet and lifestyle. Data were limited to 3 years of observation, during which time other educational and feedback mechanisms and changes in practice patterns might have influenced detection rates over time. We were also not able to adjust for experience of physicians, because the included procedures were a snapshot of the individual physician during the 3-year period. These data were not designed to call out individual physicians or outliers in secular trends. To explore this further, we would need to understand how colonoscopists value these tradeoffs and design interventions targeted at retraining these tradeoffs. Whether financial incentives play a role in improving adenoma find rates is also not known. The duration of financial interventions in this report (6 months) might not have been sufficient to demonstrate changes for this publication. Each of these questions would be important to answer in future studies. Contained in this report are several other notable findings. First, adenoma detection rates ranged from 10% 39%, despite a high completion rate of 98% and adequate preparation quality in 99% of examinations. These estimates are consistent with the literature 16 19 and also highlight that adenoma detection rate as Table 3. From GEE, Predictors of Adenoma Detection, Clustered by Individual Physicians, Adjusted for Variables Shown in Model, and Average Duration of Colonoscopy Where No Polyp Was Found OR 95% CI P value Age (y) 1.02 1.02 1.02.001 Sex (male vs female) 1.53 1.34 1.74.001 Preparation quality (adequate vs 2.26 1.64 3.12.001 inadequate) Duration of procedure a (min) 0.97 0.96 0.99.02 Sex*duration of procedure a (min) b 1.00 1.00 1.01.007 a Calculated by taking average of duration of all colonoscopies where no polyp was found by a physician, so that there is one observation per physician. b This is an interaction term of sex duration.

December 2009 VARIABILITY IN ADENOMA DETECTION AND CHANGE OVER TIME 1339 an indicator of colonoscopic quality might be independent of completion rate and adequate preparation quality. Millan et al 20 reported adenoma detection rates of 14% 27%, despite uniform completion rate of 96%. Second, variability in adenoma detection for individual physicians over time showed no systematic trends. Whether this variability has an effect on missed lesions is beyond the scope of our study and would be important for a future study. Studies are also needed to better elucidate patient and procedure-related factors that might account for the variability in adenoma detection over time. We found that increasing age, male sex, and adequate preparation quality are significantly associated with detection of polyps and adenoma. These results are similar to those reported by others. 18,21 Our finding of higher detection rate of adenoma and polyp with shorter total duration of procedure is contrary to reports from others. 22 In their prospective analysis of more than 10,000 colonoscopies, Sanchez et al 22 reported a direct correlation between total procedure time and detection of polyps (r 0.64). There are several possible explanations for this finding. First, we were unable to report withdrawal time, which is a more direct predictor of adenoma detection. 8 Second, longer total duration of procedure could imply longer time spent by physician reaching the cecum as a result of tortuous or redundant colon, and this might increase total procedure time. Third, this might reflect skill level of the endoscopist, and those with higher detection rate of adenoma and polyps might be more experienced and able to reach cecum and examine the colon in a more time-effective manner. Also, those with shorter duration also tend to do a higher volume of procedures, and this might represent a confounding variable. It should be noted that shorter duration examinations did not drive volume in this study because all examinations were scheduled for 30 minutes, and physicians were not allowed to truncate scheduled times. Furthermore, we found an interaction between sex and duration of a negative procedure, which has not been previously reported. Our finding that longer duration of colonoscopy in women might be associated with a lower adenoma detection compared with the association in men might be due to difference in level of difficulty in completing examinations in women with less than optimal visualization of the colon by the endoscopist and merits further study. Limitations of these additional findings are that we were unable to assess association of other potentially important factors such as withdrawal time and size and morphology of adenomas with adenoma detection rates. We were also unable to distinguish adenomas by histology and detection rates of advanced adenomas, which might be a more clinically relevant dependent variable. Last, although we are one of the largest community-based endoscopy practices in the country, whether our results are generalizable to the population is not known. The strengths of our study are the large sample size, averagerisk screening population, and prospectively collected data with information on adenomas. The detection of adenomas by individual physicians during a 3-year period varied and did not appear to change between individual endoscopists, despite planned, systematic interventions. This indicates that other targeted interventions might be required to improve adenoma detection rates among experienced community gastroenterologists. Future studies should evaluate variability in advanced adenoma and CRC detection rates and whether variability in adenoma detection has clinical consequences such as missed cancers. We need to establish and validate reliable quality improvement interventions to improve adenoma detection rates and understand how endoscopists make tradeoffs that determine their detection rates to be able to design interventions around them. We also need to establish adenoma and polyp detection rates in colonoscopies performed for other indications such as surveillance. References 1. Bond JH. Clinical evidence for the adenoma-carcinoma sequence, and the management of patients with colorectal adenomas. Semin Gastrointest Dis 2000;11:176 184. 2. Cotton S, Sharp L, Little J. The adenoma-carcinoma sequence and prospects for the prevention of colorectal neoplasia. Crit Rev Oncog 1996;7:293 342. 3. Rex DK. Colonoscopy: the dominant and preferred colorectal cancer screening strategy in the United States. Mayo Clin Proc 2007;82:662 664. 4. Levin B, Lieberman DA, McFarland B, et al. Screening and surveillance for the early detection of colorectal cancer and adenomatous polyps, 2008: a joint guideline from the American Cancer Society, the US Multi-Society Task Force on Colorectal Cancer, and the American College of Radiology. Gastroenterology 2008; 134:1570 1595. 5. Lieberman DA, Weiss DG, Harford WV, et al. Five-year colon surveillance after screening colonoscopy. Gastroenterology 2007;133: 1077 1085. 6. Farrar WD, Sawhney MS, Nelson DB, et al. Colorectal cancers found after a complete colonoscopy. Clin Gastroenterol Hepatol 2006;4:1259 1264. 7. van Rijn JC, Reitsma JB, Stoker J, et al. Polyp miss rate determined by tandem colonoscopy: a systematic review. Am J Gastroenterol 2006;101:343 350. 8. Barclay RL, Vicari JJ, Doughty AS, et al. Colonoscopic withdrawal times and adenoma detection during screening colonoscopy. N Engl J Med 2006;355:2533 2541. 9. Barclay RL, Vicari JJ, Greenlaw RL. Effect of a time-dependent colonoscopic withdrawal protocol on adenoma detection during screening colonoscopy. Clin Gastroenterol Hepatol 2008;6: 1091 1098. 10. Sporea I, Popescu A, Vernic C, et al. How to improve the performances in diagnostic colonoscopy? J Gastrointestin Liver Dis 2007;16:363 367. 11. Radaelli F, Meucci G, Sgroi G, et al. Technical performance of colonoscopy: the key role of sedation/analgesia and other quality indicators. Am J Gastroenterol 2008;103:1122 1130. 12. Sawhney MS, Cury MS, Neeman N, et al. Effect of institution-wide policy of colonoscopy withdrawal time or 7 minutes on polyp detection. Gastroenterology 2008;135:1892 1898. 13. Rex DK. Three challenges: propofol, colonoscopy by undertrained physicians, and CT colonography. Am J Gastroenterol 2005;100: 510 513. 14. Rex DK. Quality in colonoscopy: cecal intubation first, then what? Am J Gastroenterol 2006;101:732 734. 15. Rex DK, Bond JH, Winawer S, et al. Quality in the technical performance of colonoscopy and the continuous quality improvement process for colonoscopy: recommendations of the U.S. Multi-Society Task Force on Colorectal Cancer. Am J Gastroenterol 2002;97:1296 1308. 16. Allen JI. A performance improvement program for communitybased gastroenterology. Gastrointest Endosc Clin N Am 2008; 18:753 771, ix. 17. Rex DK. Colonoscopy practice variation. Gastrointest Endosc 2003;58:639 640. 18. O Brien MJ, Winawer SJ, Zauber AG, et al. The National Polyp

1340 SHAUKAT ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 7, No. 12 Study: patient and polyp characteristics associated with high-grade dysplasia in colorectal adenomas. Gastroenterology 1990;98:371 379. 19. Chen SC, Rex DK. Variable detection of nonadenomatous polyps by individual endoscopists at colonoscopy and correlation with adenoma detection. J Clin Gastroenterol 2008;42:704 707. 20. Millan MS, Gross P, Manilich E, et al. Adenoma detection rate: the real indicator of quality in colonoscopy. Dis Colon Rectum 2008;51:1217 1220. 21. Rex DK. Maximizing detection of adenomas and cancers during colonoscopy. Am J Gastroenterol 2006;101:2866 2877. 22. Sanchez W, Harewood GC, Petersen BT. Evaluation of polyp detection in relation to procedure time of screening or surveillance colonoscopy. Am J Gastroenterol 2004;99:1941 1945. Reprint requests Address requests for reprints to: Aasma Shaukat, MD, MPH, Division of Gastroenterology, Department of Medicine, One Veterans Drive, 111-D, Minneapolis, Minnesota 55417. e-mail: shaukat@umn.edu; fax: 612-725-2248. Conflicts of interest The authors disclose no conflicts. Funding Supported in part by grant VA Minneapolis Center for Epidemiological and Clinical Research (CECR) #04S-CRCOE-001 (A.S.) and ASGE Endoscopic Research Award (A.S.).